Email or username:

Password:

Forgot your password?
niconiconi's posts Post Back to profile
Top-level
niconiconi

Galvanic isolation turned out to be the correct decision. Just had an electrical breakdown. The SCR was completely destroyed, feeding high voltage back to the gate, then the gate drive transistor arced, high voltage kept going back further to destroy the optocoupler. ⚡💥 The sacrificial optocoupler did its job to stop damages from propagating further... #electronics

12 Jul 2022 at 13:56 | Open on cybre.space
27 comments
niconiconi

What killed my SCR? :blobcatthinking2: Either an arc over caused by low creepage & clearance, or max di/dt violation. Need to try again tomorrow... #electronics

12 Jul 2022 at 14:29 | Open on cybre.space
niconiconi

Problem solved, a n00b mistake. The SCR was incorrectly wired as a high-side switch, not a low-side one. As the load was switched on, the gate and cathode would rise to 200 volts, an exploding gate drive transistor was guaranteed... ⚡ 💥 Replaced it with a high-voltage one as a temporary fix, now it works fine. :blobfacepalm: #electronics

13 Jul 2022 at 11:03 | Open on cybre.space
niconiconi

The next long-term goal is converting this "Improvised Electronic Device" to a proper impulse generator that can produce a proper 8/20 µs surge per IEC 61000-4-5. The 4000 V / 2000 A Class 4 test is difficult to implement, but the basic 500 V / 250 A surge should be easy. #electronics

13 Jul 2022 at 14:31 | Open on cybre.space
niconiconi

Just found the full design equations for the 8/20 µs surge generator in [1]. The author was nice enough even to have all component values precalculated for you. Looks easy if you can find a suitable HV switch. I initially worried about core saturation, the surge is 100 to 1000 amps, but the L is just 10 µH, a practical value for winding an air-core coil by hand.

[1] Elementary and Ideal Equivalent Circuit Model of the 1,2/50 – 8/20 μs Combination Wave Generator, Carlo F. M. Carobbi ieeexplore.ieee.org/document/6 #electronics

Just found the full design equations for the 8/20 µs surge generator in [1]. The author was nice enough even to have all component values precalculated for you. Looks easy if you can find a suitable HV switch. I initially worried about core saturation, the surge is 100 to 1000 amps, but the L is just 10 µH, a practical value for winding an air-core coil by hand.

A power supply, charging a 5.93 μF capacitor. After the switch is closed, the high-voltage surge is discharge via a 10.9 μH inductor and 0.814 Ω series resistance to the load. Before and after the inductor, there's also a 20.2 Ω and 26.1 Ω resistors to control the surge duration.
16 Jul 2022 at 11:31 | Open on cybre.space
niconiconi

Back to switch selection... At millisecond-scale, the limitation of SCRs in an non-repetitive, oneshot pulse is thermal, as the pulse gets shorter, the allowable surge current goes up significantly, a SCR only rated for 60 A steady-state can switch even 2000 A. But at microsecond-scale, di/dt becomes the new problem, and that makes the allowed surge current to go down significantly... Apparently the responsible mechanism is localized hotspots in the silicon before it fully turns on.

Most SCRs have di/dt (max) = 50 - 100 A/μs. So for a 8/20 µs surge, the current peak must be less than 400-800 A. With a standard 2 Ω effective ESR, surge voltage is restricted to 800-1600 V. Let's see if an IGBT does this better... #electronics

Back to switch selection... At millisecond-scale, the limitation of SCRs in an non-repetitive, oneshot pulse is thermal, as the pulse gets shorter, the allowable surge current goes up significantly, a SCR only rated for 60 A steady-state can switch even 2000 A. But at microsecond-scale, di/dt becomes the new problem, and that makes the allowed surge current to go down significantly... Apparently the responsible mechanism is localized hotspots in the silicon before it fully turns on.

Figure 7. Non repetitive surge peak on-state current for a sinusoidal pulse with width tp < 10 ms. The graph shows, as the surge gets shorter and shorter than 10 milliseconds, the allowed surge current goes from 600 amps to 4000+ amps at 0.15 milliseconds. But when the surge gets shorter than 0.15 milliseconds, the allowed surge starts to go from 4000+ amps down to 300 amps at 0.01 milliseconds, because the device's di/dt limitation is 100 A/µs.
16 Jul 2022 at 14:13 | Open on cybre.space
niconiconi

IGBTs are not really much better it seems... Common discrete devices only exist below 2000 volts, roughly the same range. #electronics

16 Jul 2022 at 15:40 | Open on cybre.space
niconiconi

Pulse capacitors also become hugely expensive at 2000 volts... 2000 volts really looks like a barrier in power electronics. Beyond this point components suddenly become either hugely expensive or come with inconvenient packages... But if you stay at 1500 or 1700 V, it's much easier... #electronics

17 Jul 2022 at 23:44 | Open on cybre.space
niconiconi

Now I need 1500-volt, 6 µF pulse capacitors, what are my options.... :blobcatthinking2: It turned out the choices are pretty limited, basically three: Kemet R75 & R76 series, Vishay MKP385 series, and Cornell-Dubilier 940C & 941C series (this one seems to have a cult following by Tesla coil builders). 941C is currently the cheapest, 1.5 µF for just $11. But I need FOUR and only THREE are in stock until 2023. :blobfacepalm: #electronics

18 Jul 2022 at 1:20 | Open on cybre.space
niconiconi replied to niconiconi

High-voltage flyback transformers look like a dying breed these days. All the experimenters seem to get them either from old CRTs or random eBay surplus sources. But my project needs to be replicable from standard parts on the product catalog of a big vendor, so random components are unacceptable. CCFL lightning transformers seemed to be a good source in the past decade, but most are discontinued in this LED age. I guess I'll just stop wasting time and use a pre-made SIP DCDC module instead. #electronics

18 Jul 2022 at 19:50 | Open on cybre.space
niconiconi replied to niconiconi

The datasheet says the pulse resistors I'm using to build a 8/20 µs surge generator are themselves tested by a 8/20 µs surge generator by the vendor. Now the cycle is complete. :blobcatgiggle: #electronics

18 Jul 2022 at 20:46 | Open on cybre.space
niconiconi replied to niconiconi

You know pulse resistors are specialized when a manufacturer in China sells them at a higher price than Vishay in the US... And the official website shows zero stock. #electronics

18 Jul 2022 at 21:30 | Open on cybre.space
niconiconi replied to niconiconi

I'd just like to interject for a moment. What you’re referring to as a "8/20 μs surge", is in fact, a 1.2/50-8/20 μs surge, or as I've recently taken to calling it, the output from a 1.2/50 µs combination wave generator. The 8/20 μs waveform is not a surge unto itself, but rather just the generator's short-circuit current output as defined by IEC 61000-4-5. It's normally generated in combination with the 1.2/50 µs open-circuit voltage waveform. All the so-called "8/20 μs wave" are really the "1.2/50-8/20 μs" combinational wave... #electronics

I'd just like to interject for a moment. What you’re referring to as a "8/20 μs surge", is in fact, a 1.2/50-8/20 μs surge, or as I've recently taken to calling it, the output from a 1.2/50 µs combination wave generator. The 8/20 μs waveform is not a surge unto itself, but rather just the generator's short-circuit current output as defined by IEC 61000-4-5. It's normally generated in combination with the 1.2/50 µs open-circuit voltage waveform. All the so-called "8/20 μs wave" are really the "1.2/50-8/20...

20 Jul 2022 at 2:32 | Open on cybre.space
niconiconi replied to niconiconi

Huge pulse capacitor is huge. Cornell-Dubilier 941C series #electronics

Huge axial capacitor, 50mm x 40mm x 30mm. A single capacitor looks bigger than the circuit boards on the desk. 941C16W1P5K-F 1.5 MFD, +/- 10% 1600 VDC. CDET. KV.
20 Jul 2022 at 18:10 | Open on cybre.space
niconiconi replied to niconiconi

0.1 Ω pulse resistors arrived. Time to do more test on the original quick-and-dirty HV pulser before building a proper one: 31.2 A short circuit, 200 V open-circuit, so the effective ESR is 6.4 Ω, a bit on the high side. The di/dt is high, 10-90 rise time is just 400.0 ns, it's 65 A/µs - I'm already violating the absolute maximum of the SCR. #electronics

Oscilloscope trace, showing the load current and the source capacitor voltage.
20 Jul 2022 at 20:42 | Open on cybre.space
niconiconi replied to niconiconi

Saying I'm building a "8/20 μs impulse surge protector tester" sounds boring, I think I'll start calling it a "0.3 megawatt pulser" instead. I hope I can rightfully call it a "megawatt pulser" when I break the 2000-volt barrier one day... :blobcatgiggle: #electronics

20 Jul 2022 at 23:01 | Open on cybre.space
niconiconi replied to niconiconi

Just found a workable solution to switch 4000 volts and 2000 amps. Time to go old-school!!!

Except there are three problems. The tube has no trigger electrode, only free running is possible. Each tube costs over $200. Also, each tube is filled with 5.44 MBq of radioactive tritium ☢️ ! #electronics

The high current discharge tube GXH series double electrode spark gap is a gas discharge tube, sealed in a glass/metal casing. It can provide tubes with a DC breakdown voltage ranging from 0.50KV to 5kV. This is represented by the number or number following the letter and is expressed in hundreds of volts.
Description of High Energy Ignition Spark Gap Tubes: The GXH Series of 2-electrode spark gaps are gas discharge tubes, hermetically sealed in a glass/metal envelope. Tubes with a DC breakdown voltage in the range 0.50KV to 12 kV are available. This is signified by a numeral or numerals following the type letters and is expressed in hundreds of volts. Applications of High Energy Ignition Spark Gap Tubes: • Single-shot pulse generators • Turbine engine ignition circuits • Medical lithotripsy • High energy switches Physical Characteristics of High Energy Ignition Spark Gap Tubes: All ratings given in this data sheet a reab solute, nonsimultaneous ratings. It is the equipment designer’s responsibility to ensure that they are not exceeded. The spark gap life depends on circuit conditions such as peak discharge current and duration, charge transfer per discharge and the repetition rate. Charge transfer, single discharge (8/20 ms waveshape): 0.5C max Cumulative charge transfer: 6000C at 0.81 mC per discharge at 1 Hz Operating temperature: -55 to +125 C. Mechanical shock, half-sine (to B.S. 2011 Es): 40g for 4ms Mounting: screw mounted Radioactivity: tritium (3H), 5.44 MBq (147 mCi) max per device.
21 Jul 2022 at 7:28 | Open on cybre.space
niconiconi replied to niconiconi

Just finished the construction of my new 8/20 μs impulse generator. Its output surge should be in compliance with IEC 61000-4-5 now, but I haven't checked it yet. Need to get some sleep and continue the test tomorrow. Working with lethal voltage at midnight is definitely not a good idea... #electronics

24 Jul 2022 at 16:21 | Open on cybre.space
niconiconi replied to niconiconi

My impulse generator's current output failed to meet the standard waveform requirement. The rise time is 50% too short. I think I didn't wind that inductor correctly. Just ordered a simple LCR meter to continue the experiment (trimming an inductor with a oscilloscope or VNA is just too awkward). #electronics

25 Jul 2022 at 8:24 | Open on cybre.space
niconiconi replied to niconiconi

I found the solution for switching 2000 volts - two thyristors in series should be able to handle it. I was worriying about off-state leakage current due to voltage balancing resistors in the original capacitor discharging circuit. But this problem disappears in the standard 8/20 μs impulse generator, it already has two 20 Ω resistive loads, this diverts leakage mostly to the ground instead of the load, making the series switch connection feasible again.

Capacitor size and cost are still the problem. Using non-pulse rated capacitors is perhaps an acceptable workaround if I'm willing to replace them once in a while. A lifetime of 10,000 shots is practically infinite for my purpose. #electronics

I found the solution for switching 2000 volts - two thyristors in series should be able to handle it. I was worriying about off-state leakage current due to voltage balancing resistors in the original capacitor discharging circuit. But this problem disappears in the standard 8/20 μs impulse generator, it already has two 20 Ω resistive loads, this diverts leakage mostly to the ground instead of the load, making the series switch connection feasible again.

26 Jul 2022 at 16:04 | Open on cybre.space
niconiconi replied to niconiconi

LCR meter arrived, indeed, the inductor I winded was 50% too low. I pushed the windings together while looking at the readings, within 10 seconds I brought it within design specification.

Time for a test... Then I immediately blow the entire circuit up by connecting 200 V to 5 V input... ⚡💥 :doge: #electronics

27 Jul 2022 at 10:54 | Open on cybre.space
niconiconi replied to niconiconi

Success, the output waveform of my 8/20 μs impulse generator prototype is now IEC 61000-4-5 compliant!

Peak current: 200 A
Front time: 7.05 μs
Duration: 21.24 μs #electronics

A homebrew perfboard with four huge HV capacitors, some large HV power resistors, an air-core inductor coil, and a TO-220 thyristor at the center. At the left, there's a big red button, galvanic isolated from the HV power domain, for triggering the surge. At the leftmost side is a ready-mode HV power supply module.
The standard 8/20 μs impulse current waveform in IEC 61000-4-5. [Figure 3 – Waveform of short-circuit current (8/20 µs) at the output of the generator with no CDN connected] Front time: T f = 1,25 × Tr = 8 µs ± 20 % Duration: T d = 1,18 × Tw = 20 µs ± 20 % Undershoot: 0% to 30%. NOTE 1 The value 1,25 is the reciprocal of the difference between the 0,9 and 0,1 thresholds. NOTE 2 The value 1,18 is derived from empirical data.
Oscilloscope measurement of the output waveform of the actual surge generator. Rise time is 5.363 μs, peak current 194 A, no undershoot.
Same oscilloscope trace with cursor measurement of time interval between the instant at which the surge voltage rises to 0,5 of its peak value, and then falls to 0,5 of its peak value. Duration is 18.00 μs.
27 Jul 2022 at 14:34 | Open on cybre.space
niconiconi replied to niconiconi

The only problem I need to solve now is that the entire circuit exploded after firing the first shot. ⚡💥 I believe it was the same mistake and failure mode - the thyristor was wired as a high-side switch this time deliberately for convenience. As soon as the thyristor is turned on, the gate voltage rises to 500 volts.

I thought the transistor was protected by the new diode I added in series, but no, it's again a n00b mistake. The base-collector junction of a BJT is also a diode, and you can't connect any diodes in series without an RC snubber to balance the voltage. Without balancing, series diodes simply break down one after another. #electronics

The only problem I need to solve now is that the entire circuit exploded after firing the first shot. ⚡💥 I believe it was the same mistake and failure mode - the thyristor was wired as a high-side switch this time deliberately for convenience. As soon as the thyristor is turned on, the gate voltage rises to 500 volts.

27 Jul 2022 at 15:00 | Open on cybre.space
niconiconi replied to niconiconi

New plan: abusing a MOSFET gate driver to drive a SCR's gate to get high-side switching. Now I need to figure out how much Common Mode Transient Immunity do I need. My circuit is *literally* a surge generator. #electronics

27 Jul 2022 at 17:17 | Open on cybre.space
niconiconi replied to niconiconi

CMTI: 100 kV/μs. Oh, it's more than enough... #electronics

27 Jul 2022 at 17:21 | Open on cybre.space
niconiconi replied to niconiconi

Rewired the isolated DC-DC converter to float on top of the high voltage instead of ground. No more explosions, even a small-signal transistor can drive the gate with ease. #electronics

29 Jul 2022 at 14:31 | Open on cybre.space
niconiconi replied to niconiconi

Look at my gate turn-on waveform... Man-made horrors beyond comprehension. #electronics

A step signal, with extremely large overshoots and ringings for several microseconds at the rising edge.
29 Jul 2022 at 14:35 | Open on cybre.space
niconiconi replied to niconiconi

In electronics, everything is an LC resonator if your layout is bad enough. #electronics

29 Jul 2022 at 14:38 | Open on cybre.space
Go Up